On Mon, Jan 25, 2010 at 01:27:19PM -0200, Luiz Capitulino wrote: > On Mon, 25 Jan 2010 15:35:53 +0100 > Markus Armbruster <arm...@redhat.com> wrote: > > > Luiz Capitulino <lcapitul...@redhat.com> writes: > > > > > On Mon, 25 Jan 2010 12:09:06 +0200 > > > "Michael S. Tsirkin" <m...@redhat.com> wrote: > > [...] > > >> Finally, > > >> don't we want unsigned values in protocol? > > > > > > JSON doesn't support them. > > > > Uh, where does the RFC say that? > > I see that my comment was misleading. > > In JSON we don't have unsigned types, we have only a type > called 'number' to represent them all. > > Unsigneds should be handled correctly, except for uint64_t which > is cast to int64_t. > > Michael, does this answer your question? > Is there any > issue with the handling of unsigneds I'm not aware about?
The issue I see isn't related to unsigned. Apparently we currently accept values such as 'a' as valid strings. Since this is not valid json we probably should reject it just in case we will want to switch to another json library, otherwise clients might come to depend on non-standard behaviour. -- MST
